Skip to main content
Internet Archive's 25th Anniversary Logo

Full text of "The Signature of Primordial Grain Growth in the Polarized Light of the AU Mic Debris Disk"

See other formats

ApJ in press 

Preprint typeset using I^T^^ style cmulatcapj v. ^jllj^A, 



James R. Graham^' Paul G. Kalas^'^, 


Brenda C. Matthews 

ApJ in press 


We have used the coronagraphic mode of the Advanced Camera for Surveys aboard the Hubble Space 
Telescope to make the first polarization maps of the debris disk surrounding the nearby M star AU 
Microscopii (GJ 803). The hnear polarization of the scattered light from the disk is unambiguously 
detected. We find that the degree of polarization for the disk rises monotonically from about 0.05 to 

0. 35 between projected radii of 20 and 50 AU. Polarized light is detectable out to about 80 AU, where 
the fractional polarization reaches a maximum observed value of 0.41 ± 0.02. Polarization vectors 
are oriented perpendicular to the disk mid-plane, indicating that polarization originates from single 
scattering in an optically thin dust disk where the albedo is dominated by small {x — 27ra/A < 1) 
particles. We use simple, optically thin disk models to infer the spatial structure of the disk and the 
scattering properties of the constituent grains by simultaneously fitting the surface brightness profile 
and the degree of linear polarization. The best fit models require that the dust grains exhibit high 
maximum linear polarization and strong forward scattering. The inner disk (<40-50 AU) is depleted 
of micron-sized dust by a factor of more than 300, which means that the disk is collision dominated, 

1. e., grains that are dragged inward by corpuscular and Poynting- Robertson drag undergo a destructive 
collision. While the inferred optical properties are covariant with the radial distribution of dust, the 
only acceptable models have Pmax > 0.50 and g > 0.7. These constraints cannot be met by spherical 
grains composed of conventional materials. A Mie scattering analysis implicates grains where the 
real part of the refractive index is about 1.03, which is a signature of highly porous (91-94%) media. 
More reliable methods for calculating the scattering properties of aggregates confirm that the observed 
values of Pmax and g can be matched by high porosity, micron-sized clusters of small particles. In 
the inner Solar System, porous particles form naturally in cometary dust, where the sublimation of 
ices leaves a "bird's nest" of refractory organic and silicate material. In AU Mic, the grain porosity 
may be primordial, because the dust "birth ring" lies beyond the ice sublimation point. The observed 
porosities span the range of values implied by recent laboratory studies of particle coagulation in the 
proto-solar nebula by ballistic cluster-cluster aggregation. To avoid compactification, the upper size 
limit for the parent bodies is in the decimeter range, in agreement with theoretical predictions based 
on coUisional lifetime arguments. Consequently, AU Mic may exhibit the signature of the primordial 
agglomeration process whereby interstellar grains first assembled to form macroscopic objects. 

Subject headings: polarization — dust — stars: individual(GJ 803; AU Mic) — planetary systems: 
formation — circumstellar matter 


Approximately 15% of nearby main sequence stars 
manifest infrared excess due to the reprocessing of stel- 
lar radiation by dust grains in a circumstel lar disk 
()Aumann et al.lll984t iBackman fc Parescelll993|) . These 
systems are known as "debris disks" because the lifetime 
of dust is orders of magnitude shorter than the stellar 
age, suggesting a continuous supply of fresh grains re- 
leased from larger, undetected parent bodies. In the so- 
lar system, interplanetary dust particles are resupplied 
by the coUisional erosion of asteroids and the sublima- 

^ Astronomy Department, University of California, Berkeley, CA 
94720-3411, U.S.A. 

^ Center for Adaptive Optics, University of California, Santa 
Cruz, CA 95064, U.S.A. 

* Herzberg Institute of Astrophysics, National Research Council 
of Canada, 5701 West Saanich Road Victoria, BC, Canada V9E 

tion of comets. Cometary dust grains retain the his- 
tory of their interstellar origin, but are imprinted with 
structures that speak to their incorporation into larger 
bodies and subsequent return to interplanetary space 
([Greenberg & Hage 1990). Therefore, cometary dust 
provides our closest link to the particle coagulation pro- 
cesses that occurred during the earliest phases of planet 
building in the Solar System. Although debris-disk dust 
must be highly processed and modified during incorpora- 
tion into, and subsequent attrition of, large bodies, these 
particles are our primary source of information regarding 
the growth of assembly of planetesimals in exoplanetary 
systems. The laboratory study of low-velocity dust in- 
teractions thought to be characteristic of conditions in 
the early solar nebula suggests that particles grow un- 
der ballis tic cluster-cluster ag,g regation into fractal as- 
semblies ()Wurm fc BlumllT998jr . These bodies suffer re- 
structuring when the aggregate diameters exceed a few 
centimeters. Smaller bodies are not expected to be sub- 


Graham et al. 

jected to impact compaction l|Blum fc Wnrmll2(ffl(il) . As 
these clusters have unique optical properties, the study 
of debris disks provides experimental validation of our 
ideas regarding growth of soli d bodies from inters tellar 
grains to macroscopic objects ( Kimura et alJl20?)fi^ . 

At visible wavelengths debris-disk dust can be detected 
in scattered light, analogous to the Zodiacal light in our 
Solar System. As in the Solar System, the polarization 
state of this scattered light is a key diagnostic of the 
grain properties. In an optically thin disk, where single 
scattering dominates, a high degree of linear polariza- 
tion is expected, with a characteristic orientation that 
is symmetric (c oncentric or ra. dial) about the illuminat- 
mg source fe.g.. lKruegeill2003j) . The observed de gree of 
polarization is an important indicator of the size, shape, 
composition, physical structure and alignment of individ- 
ual grains and their distribution along the line of sight. 
Single particle scattering can be described by the matrix 
elements of the complex amplitude scatte ring function, 
S, w hich depend on the scattering angle ijvan de HulstI 
I1981D . The angular dependence of 5 is a key clue to 
the nature of the particles. For example, the degree of 
forward scattering increases with particle size from the 
Rayleigh limit {x — 27ra/A <C 1), where the scattering 
asymmetry parameter, g — {cos0) ~ 0, to 5 ~ 1 as x ap- 
proaches unity. Observations of spatially resolved disks 
can directly determine the azimuthal asymme try due to 
asymmetric scattering fe.g.. lKalas et al.ll2005j) . Because 
debris disks are optically faint relative to their illuminat- 
ing star, there is a strong selection effect that favors the 
discovery and facilitates the study of high surface bright- 
ness, edge-on systems, e.g., /3 Pic, HD 32297, HD 139664, 
or AU Mic. 

In an edge-on system, the azimuthal variation of sur- 
face brightness around the star is unobservable, and thus 
grain scattering properties versus phase angle cannot be 
measured directly. Each line of sight through the disk in- 
cludes a range of scattering angles. Therefore, both the 
asymmetry parameter and the radial variation of dust 
opacity determine the disk surface brightness as a func- 
tion of angle on the sky. Unless there is prior information 
regarding g or the dust distribution these properties can- 
not be uniquely disentangled from measurements of the 
radial surface brightness distribution. Inclusion of fur- 
ther constraints, such as the spectral energy distribution 
(e.g.,[Strubbc & Chiane 200|, can break this degener- 
acy. Polarization data can also play this role. In general, 
the matrix elements of S have different angular depen- 
dences. Therefore, observation of the intensity, which 
depends on jS'ip -I- 15*21^, and the polarization state of 
scattered light, which depends on jS'ip — 15*2^, can be 
used to recover this otherwise lost information. 

Here we present the first optical polarization study of 
the AU Mic debris disk. AU Mic is a nearby (9.9 pc) 
dMle star with Galactic space velocities that suggest a 
comm on origin with P Pic. (Barrado v Navascucs ct al\ 
II999V The discovery of scattered visible light from a 
near-edge-on debris disk around AU Mic (^Kalas et al.l 
|2004|) supports the picture of /3 Pic, AU Mic and nearly 
20 other stars as a coeval group with age I2I4 Myr 
(jZuckerman et aL|[2bOL'l . The /3 Pic debris disk has been 
studied extensively, including the finding of polarization 
along the mid-plane varying between 0.12 and 0.21 in op- 

tical data ijGledhill et al.ll991l:lm)lstencroft et al.ll995(l . 

These meas urements have been interpreted by several au- 
thors (e.g.. | Art vmowic 3 l l997t iVoshchinnikov fc: Kriigel 
Il999t iKrivova et al. 200(3). recent ground based ob- 
servations^^ polarization in the K band has been detected 
ijTamuraet a l..200fi). 

Despite the l arge difference in stellar mass, 

iStru bbe k Chiand l)2005j) and lAugereau fc BeustI 
{200&) argue that the radiation pressure force that 
quickly expels small /3 Pic grains has a counterpart 
around AU Mic in the form of stellar wind. To first 
order this explains the initial finding that the disk 
mid-plane surface brightness dis tribution is nearly 
identical for both debris disks iKalas et alJ |2004). 
However, t he blue color gradient for AU Mic's mid-plane 
llMetchev et al. 2005: Krist ct al. 2005; Fitzgerald ct al] 
2006 ) contrasts aga inst the red color of (3 Pic's mid-plane 
l|Golimowski et al]f2006;l . This distinct color difference 
between the two disks points to a significant divergence 
in either the grain composition, grain size distribution, 
or minimum grain size. Polarization observations of AU 
Mic are therefore valuable in constraining these grain 
properties and identifying the fundamental differences 
between the two disks. 

Our purpose is to present Advanced Camera for Sur- 
veys (ACS)/High Resolution Camera (HRC) polarime- 
try of the AU Mic system and to shed some initial 
light on the disk structure and grain optical proper- 
ties that are emphasized by the detection and measure- 
ment of the disk in polarized light. The fidelity of any 
model increases with the observational challenges pre- 
sente d to it, thus it must fit the near-IR and optical emis- 
sion ( Kalas et alJl200i iLhH I200I iMetchev et aP [200I 
Fitzg erald et al.ll2006[) and the IR to sub-mm spectral 
energy distribution l|Liu et all 120041: IChen et al.ll2005^1 . 
We defer this synthesis to a later study. 

Section [3 outlines the observations, point spread func- 
tion (PSF) subtraction, calibration of the Stokes param- 
eters, and correction for instrumental and interstellar 
polarization. Section reports the appearance of the 
AU Mic disk in polarized light. Section [3.21 provides a 
qualitative description of the extracted one-dimensional 
surface brightness profile and degree of linear polariza- 
tion, compares AU Mic with /3 Pic, and draws some 
preliminary conclusions regarding the radial distribu- 
tion of the dust. We describe quantitative analysis us- 
ing optically-thin edge-on disk models in ^ and develop 
a method that simultaneously fits the observed surface 
brightness and fractional polarization to three dust mod- 
els: a semi-empirical Henyey-Greenstein model, a Zodi- 
acal dust model, and a Mie model. The successes and 
failures of these models are addressed in SjSl and the evi- 
dence for porous grains is examined. Our conclusions are 
summarized in SjHl 


Coronagraphic observations of AU Mic (GJ 803, HD 
197481) were made on 2004 August 01, using the 1'.'8 di- 
ameter (~ 64A/D at V) occulting spot on the ACS/HRC 
aboard Hubble Space Telescope (HST). The occulting 
spot is located in the aberrated beam from HST, be- 
fore correctiv e optics, and int ercepts about 88% of the 
on- axis light l)Ford et al.ll2003D . The F606W broadband 

Polarization of AU Mic 



- 1 1 % 

- ea 

: O 

O 10 PC a - 
„ 100 PC O 
W 1000 PC O 

o O Q ® ' 

i o ^' 

■ o e 

<2r o : 


10 20 
I [deg] 



Fig. 1. — Interstellar polarization for stars in the neighborhood 
(< 25 degrees) of AU Mic |Hcilos 2000). Stars are plotted in Galac- 
tic coordinates, centered at the location of AU Mic (intersecting 
dotted lines). The length of the lino designates the degree of lin- 
ear polarization, and the diameter of the circle is proportional to 
the logarithm of the distance. The vertical tick mark labeled 1% 
gives the degree of polarization scale. The median measured po- 
larization is 0.1% in this direction suggesting that the interstellar 
polarization of AU Mic is small and can be neglected. 

filter, comparable to the Johnson-Cousins V band, was 
chosen for all observations. AU Mic was observed during 
two orbits with the spacecraft roll angle offset by 10.1 
degrees between orbits. Three exposures of 240 seconds 
were obtained through each of the three polarizer ele- 
ments (POLOV, POL60V, POL120V). A third orbit was 
devoted to observing a PSF reference star using the same 
filter combinations. Together, PSF-subtracted, corona- 
graphic data render an improvement factor exceeding one 
hundred in contrast r elative to direct im aging. The PSF 
star (GJ 784; MOV l|Evans et al.lll957|) ) was chosen on 
the basis of similar brightness to AU Mic, close spectral 
type match, and proximity on the sky. Data reduction 
included the standard pipeline processing from the HST 
archive that produces bias-subtracted and flat-fielded im- 
age files. Images were additionally processed using the 
recommended spotflat and pixel area map. Differencing 
frames within an orbit and between orbits was used to 
register all images. 

PSF subtraction at each POLV filter was achieved by 
subtracting the PSF reference star from AU Mic to pro- 
duce a residual image that gives a mean radial profile 
equal to zero intensity in directions perpendicular to 
the mid-plane. PSF subtraction was implemented be- 
fore construction of the Stokes parameter images be- 
cause the POLV filters introduce filter-specific artifacts. 
The polarizer filters contribute a weak geometric dis- 
tortion which rises to about 0.3 pixels near the edges 
of the HRC. This is caused by a weak positive lens in 
the polarizers, which is needed to maintain proper fo- 
cus when multiple filters are in the beam. In addi- 
tion, the visible polarizer has a weak ripple structure 
which is related to manufacture of its Polaroid material; 
this contributes an a dditional zt 0.3 pixel distortion with 
a complex structure llBiretta fc Kozhurina-Plataijl2004l: 
iKozhurina-Platais fc Birettall2004l) . All these geometric 
effects are correctable, but astrometry obtained with the 
POLV filters will likely have reduced accuracy due to 
residual errors and imperfect corrections. 

2.1. Calibration of the Stokes Parameters 

Imaging p olarimetry with ACS is d escribed and char - 
acterized bv'B iretta et al.l l)2004|) and iPavlovskvl (^006). 
The ACS/HRC polarimeter is implemented as three an- 
alyzers installed in a filter wheel at nominal angles of 
0, 60 and 120 degrees denoted POLVO, POLV60, and 
POLV120, respectively. Mueller matrix algebra can be 
used to show that the observed intensity, Ig, through a 
perfect analyzer rotated by an angle 9 is 



and U are the Stokes parameters 
111113). This arrangement is insensitive 


to the circular polarization, V, which we assume hence- 
forth to be negligible. Assuming that the three POLV 
filters are ideal, and ignoring instrumental polarization, 
we can solve the set of resultant simultaneous equations 
to show 

^ — g (^0 + -^60 + -^120) 

Q = 2 (2^0 ^ ^60 ^ -^120) 


We measure instrumental angles counter-clockwise from 
the +Q axis, which we take to be defined by the ori- 
entation of the POLVO filter. We quote this elemcntar 
result because Biretta et al. ( 2004) and Pavlovskv ( 200' 
each use different sign conventions for angles, resulting 
in different expressions for U; our choice is consistent 
with the latter. The Stokes parameters are then pro- 
jected onto an astronomical coordinate system using the 
Mueller matrix to rotate from the spacecraft reference 
frame. The degree of linear polarization, which is oc- 
casionally referred to as the fractional polarization, is 
defined as p = 

(Q2 + U'^f/^/I. The position angle of the 
electric field is -0 = (1/2) arctan([//Q). 

The polarizing efficiency of the POLV filters is high: 
{Ts - Tp)/{Ts + Tp) > 0.999 at 600 nm, where T is the 
transmission for the s and p linear polarization states, 
respectively, and the orientation of each POLV filter 
is within a degree of its nominal values ijBiretta et alJ 
2004; Biretta & Kozhurina-Platais 2004). Nonetheless, 
the ACS is not an ideal polarimeter. The HRC em- 
ploys three non-normal reflections and a tilted detec- 
tor that combine to yield an instrumental polarization 
in F606W of 0.063 at a position angle oi tp ^ —87 de- 
grees^ (,Biretta et al.,..2004.') . 

Using our sign convention. 


Graham et al. 

Fig. 2. — A pseudocolor representation of the AU Mic debris disk in Stokes / measured with ACS/HRC in F606W (Ac= 590 nm, AA = 
230 nm). Over-plotted are ticks that indicate the orientation of the electric field. The length of the tick is proportional to the degree of 
linear polarization. A 50% polarization tick is indicated. The E-field vectors are derived from Stokes parameters that have been binned x8 
into 0'.'2 pixels prior to calculating the degree of polarization. The binned vectors are fully independent. The degree of polarization rises 
smoothly from about 5% close to the star up to approximately a peak linear polarization of 40%. The high degree of polarization, and the 
orientation of the electric vector perpendicular to the disk plane are indicative of small-particle scattering in an optically thin disk. Data 
within a radius of about I'.'O suffer from significant systematic errors due to imperfect PSF subtraction. 

We remove the instrumental polarization from our data 
by using the co rrection factors, which are derived by 
iPavlovskvl l)2006|) from observations of polarization stan- 
dard stars. These corrections are applied to the observed 
count rate in each of the three polarizers before comput- 
ing the Stokes parameters. The systematic error in the 
degree of linear polarization for a weakly polarized source 
is about ± 0.01; the resultant degree of hnear polariza- 
tion will have a fractional systematic error of about 10% 
for highly polarized {p > 0.20) sources. For example, the 
systematic error for p = 0.05 is 0.01, but the systematic 
error for p = 0.50 is 0.05. The systematic error in tp is 
about 3 degrees. Because we combine two observations 
at different roll angles, our systematic errors are some- 
what reduced relative to these values. 

The y-band polarization standards BD -1-64° 106 {p — 
569 ± 4 X 10-4; ip = 96.6 ± 0.2 d egrees) and GD 319 
(p = 9 ± 9 X 10-") l|Schmidt et alJ ll992,) were observed 
with the HRC, F606W, and the POLV filter set, as part 
of the ACS polarization calibration program (HST Pro- 
posal IDs 9586 & 9661). We established the correctness 
of our implementation of the Stokes parameter calibra- 
tion procedure by measuring the degree of polarization 
and PA of these stars, comparing with the ground-based 
results, and confirming agreement within the statistical 
errors. The polarization calibration targets are weakly 
polarized point sources, whereas the AU Mic disk is spa- 
tially extended. To gain experience and confidence with 
imaging polarimetry we also analyzed the ACS Wide 
Field Camera F606W/POLV observations of the highly 
polarized Crab Pulsar synchrotron nebula (Proposal ID 

2.2. Interstellar Polarization 

In additional to correction for instrumental polariza- 
tion observations should also be corrected for interstellar 
polarization. AU Mic is nearby, at high Galactic latitude 
and has minimal color excess, suggesting that polariza- 
tion due to interstellar dust grains is likely to be neg- 
ligible. Although the polarization of AU Mic itself has 
not been detected in integrated light ijPettersen fc Hstil 

I1981() . we can inspect the interstellar polarization of ad- 
jacent stars (Fig. This figure shows the observed 
degree of linear polarization for star s within an a ngular 
radius of 25 degrees of AU Mic from lHeilesI l|2000(l . The 
median measured polarization is 1 x 10-'^ in this direction 
on the sky and the highest observed value is 8 x 10-^, and 
this is for a star considerably more distant (260 pc) than 
AU Mic. We therefore neglect any interstellar polariza- 
tion in the subsequent discussion. AU Mic is a flare star, 
and may exhibit flare-generated particle b eam polarized 
light l|Kundu et alJll987t ISaar et alJll994l) . 

3.1. Linear Polarization of the AU Mic Disk 

Stokes /, Q and U images were created from the PSF- 
subtracted POLVO, POLV60 and POLV120 images as 
outlined in t|2.1l The corresponding degree of linear po- 
larization and the orientation of the electric field are rep- 
resented in Figure [3 Because the Stokes parameters are 
formed from a linear combination of the observed flux, 
their statistical properties are simple. In contrast, the 
degree of polarization is a positive definite quantity, and 
therefore biased. Consequently, the polarization infor- 
mation that we display in Figure |3 has been derived from 
well binned Stokes images prior to computation of p and 

to ensure that this figure gives a reliable impression of 
the results. 

Figure|21shows that the degree of polarization increases 
monotonically with distance from the star from about 
0.05 to 0.40 (see also Fig. O. Everywhere, the electric 
field is consistent with an orientation perpendicular to 
the disk. These two results are qualitatively in accord 
with the expected signat ure of sca ttering by small spher- 
ical particles with x <1 ijKrueeel ' 2003). It is likely that 
the variation of polarization with distance from the star 
occurs because, along a given line of sight, a range of 
scattering angles contributes to the observed intensity. 
The degree of linear polarization typically peaks at scat- 
tering angles close to 7r/2, and polarization is always zero 
in the forward and backward directions. Thus, the peak 
polarization signal is diluted by light arising from more 

Polarization of AU Mic 



£. 1000 r 

Fig. 3. — One dimensional Surface brightness in Stokes / (top) 
and degree of linear polarization (bottom) as a function of pro- 
jected radius, b, for AU Mic. The surface brightness is in units of 
detected photoelectrons per 0''025 (0.25 AU) wide pixel column. 
Errors represent statistical errors only. Systematic errors are in- 
cluded in Figure ISl 

acute and more oblique scattering events. We expect to 
see the peak polarization at the outer edge of the disk, 
where only right-angle scattering contributes. If the disk 
is devoid of dust within some inner boundary ri, then 
for impact parameters b < ri scattering events with an- 
gles between arcsin(&/ri) and tt — arcsin(6/ri) are absent, 
and the degree of polarization is reduced even further 
(see Figure EJ. For an optically thin disk, where single, 
small-particle scattering dominates, the electric field is 
oriented, as observed here, perpendicular to the plane 
containing the star, the dust grain and the observer. For 
intermediate size (a; ^ 1) spherical particles the plane 
of polarization can flip by 7r/2 at certain scattering an- 
gles so that the electric field i s oriented parallel to the 
scattering plane l)Kruege]|l2003j) , which is clearly not the 
case here. Moreover, particles composed of conventional 
astrophysical grain materials show large amplitude os- 
cillatory behavior in \Si\'^ — 15*21^ with scattering angle, 
with angular period 59 ~ A/2a. Thus any line of sight 
that comprises emission from a range of scattering an- 
gles, A6', such that the particle size satisfies A9/56 ^ 1, 
will tend to exhibit weak linear polarization. 

3.2. One- dimensional Profiles 

We have measured one-dimensional surface brightness 
profiles along the disk in /, Q and U for comparison 
with simple disk models. This photometry was extracted 
optimally using column-by-column fitting to the verti- 
cal surface brightness profiles. This approach is advan- 
tageous because the mid-plane location and projected 
disk thickn ess vary significantly with impact parameter 
ijKrist et a l. 2005). Figureglshows the full-width at half 
maximum (FWHM) thickness of the disk (in units of 
the impact parameter, &), as a function of b. Between 
projected radii of 25 and 50 AU the FWHM grows rel- 
atively slowly as ~ then beyond 50 AU the thick- 
ness increases more rapidly as ~ b^^^. Because of this 
variation, it is unsatisfactory to simply measure the disk 
signal in an aperture of fixed height: a varying fraction 
of the emission is missing from a small aperture, while 
excess noise contaminates large apertures. Neither an 
exponential nor a Gaussian describe the vertical profile 


b [AU] 

Fig. 4. — Vertical thickness of the disk (in units of b) derived 
from F606W Stokes 7 as a function of projected separation, b. 
The projected FWHM of the disk is measured by fitting a Cauchy 
function (Eq. 0, where 2h is the FWHM. The strong variation 
of disk thickness means that photometry in a fixed aperture is not 
an effective means for extracting the 1-d surface brightness profile. 
The vertically integrated surface brightness profile shown in Fig. |21 
is simply C{b). The disk thickness varies with projected separation, 
and shows two distinct regimes: within 50 AU h ~ fe^/^ and beyond 
50 AU h ~ fe^/^. The thin lines show robust least-squares fits of 
the form h/b^b°' for 20 < b/AU < 50 (soUd) and 6 > 50 (dashed). 

adequately. However a Cauchy distribution 

c{b,z) = C{b) 



with FWHM, 2h, provides an excellent fit at all impact 
parameters, where z is the coordinate perpendicular to 
the disk plane, and zq is the location of the mid-plane. 
The 1-d surface brightness profile is then simply the val- 
ues C(6) = / c{b, z) dz derived from this fit. The data 
displayed in Figure 0] are derived from fitting the Stokes 
/ image. Because the surface brightness declines with 
radial distance, we have binned the vertical profile in in- 
creasingly wide blocks with distance from the star. The 
data are binned into three-pixel (0'.'075) wide columns for 
r < 30 AU increasing to nine-pixel (0'.'225) wide columns 
for 50 < r/AU < 70 AU. This binning reduces the un- 
certainties at the outer edge of the disk at the expense of 
lower angular resolution. The signal-to-noise is greater 
in Stokes / than in Q or U . Therefore, we assume that 
the disk thickness and mid-planc location docs not vary 
with the polarization state and use the results from fit- 
ting Stokes / to constrain the Q and U fits. The resultant 
Stokes / 1-d surface brightness profile and degree of lin- 
ear polarization is shown in Figure |3| 

The polarization measurements of (3 Pic's debris disk 
provide a natural point of reference for comparison with 
our results. A qualitative consideration of the factors 
described in H3. II suggests that these two disks have dif- 
ferent polarization signatures because they have different 
radial dust distributions, and the two disks are measured 
on different scales. The i?-band polarization of ( 3 Pic's 
disk is observed to range from 12 to 0.17 (jGledhill et alJ 
119911: IWolstencroft et aO 11995(1 : the degree of polariza- 
tion shows a weak gradient, increasing outward between 
200 and 600 AU. In contrast, polarization of the AU Mic 
disk rises quickly by a factor of about five between pro- 
jected radii of 20 and 50 AU. A direct comparison of AU 
Mic with (3 Pic is not possible because the spatial scales 


Graham et al. 

Fig. 5. — Model disk geometry. The surface brightness at impact 
parameter b is evaluated along the line of sight, £. The scattering 
cross section is a function of the scattering angle, 9 = arcsin(fe/r). 
It is convenient to make the change of variable in Eq. to 9, 
in which case the limits of integration become arcsin(fe/r2) and 
TT — arcsin(b/r2). 

probed do not overlap — partly because /3 Pic at 19.3 pc 
is almost twice as distant as AU Mic, and partly because 
the P Pic measurements are derived from seeing limited 
observations. Nonetheless, the differences suggest that 
the rapid rise of the linear polarization of AU Mic's disk 
between 20 and 50 AU occurs because these lines of sight 
intersect a central hole where scattering angles ~ 7r/2 
are absent. In f3 Pic this steep rise in unobserved, be- 
cause the dust depleted zone lies too close to the star to 
be readily observable from the ground. Recent near-IR 
adaptive optics data that probe /3 Pic's debris disk on 
scales of 60- 120 AU are consiste nt with a 120 AU inner 
hole radius ijTamura et alJl20?)^ . Th e visible extent of 
the Pic disk is at least 1800 AU ijLarwood fc KalasI 
1200 ID . Evidently, ri/ra is smah (~ 0.1) for P Pic, and 
assuming that the optical properties of grains are ho- 
mogeneous across this disk then the polarization should 
continue to increase gradually with increasing impact pa- 
rameter out to the outer edge, located at r2. 

This explanation, which invokes only geometric factors 
to explain the difference between AU Mic and f3 Pic, is 
incomplete on two counts. First, the outer radius cannot 
be ~ 100 AU, because the AU Mic disk is traced out to 
210 AU l|Kalas et all 12004(1 . Second, the peak detected 
polarization beyond 50 AU exceeds 0.30, which exceeds 
the peak linear polarization of the dust invoked to ex- 
plain the P Pic measurements. Taken together these two 
observations imply that the peak linear polarization of 
an individual scatterer in the AU Mic disk must exceed 
0.40. However, a quantitative comparison in 21 shows 
that our inferences about the relative scales of these two 
disks contain a grain of truth. 


The linear polarization of /3 Pic's disk can be explained 
by assuming a radial, power- law opacity distribution and 
the optical properties of Solar System dust grains, ex- 
amples of which include the Zodiacal-light grains, in- 
terplanetarv dust p articles (IDPs), and cometary dust 
(|Artvmowiczl Il997fi . It is therefore useful to enquire 
whether the polarization signature measured in tl3.2l can 
be described by such a model, and whether the differ- 
ences between AU Mic and /3 Pic f H3.1|) can be attributed 
solely to different radial opacity distributions. 

Suppose that the surface density of dust follows 
a power law, and the grain properties are uniform 
throughout the disk, such that the vertical optical 
depth to scattering presented by the grains is T±{r) = 
J__n{r, z)Tra^Qsca dz = Tj^,i(ri/r)''. The differential 
scattering cross section per unit solid angle of each dust 
grain is na^QscaSiiO), where a is the geometric grain ra- 
dius and Qsca is the scattering efficiency. We use i to 
denote either the s or p polarization states (with respect 
to the scattering plane) , and 9 indicates the dependence 
of the cross-section on the scattering angle. The phase 
function, Si, is normalized so that J Si d£L = 1. For 
anisotropic particles, the cross-sections are functions of 
9 and 0. We assume isotropy, and ignore the azimuthal 
dependence, i.e., the grains are spherical, or randomly 

The observed one-dimensional surface brightness (cf. 
Eq. lO) of an optically thin disk is expressed by an 
integral along the line of sight, I, 

C,{h)=L, -^s,{9)di, (6) 

where is the monochromatic luminosity of the star at 
the observing frequency, r is the radial coordinate, and 
ri and r2 are the inner and outer radii of the disk (see 
Figure \^ . This integral can be rewritten by change of 
variable to the scattering angle using d9 = —b d£ and 
setting the limits of integration to 02 = arcsin(&/r2) and 
TT — 02- Evaluation of the surface brightness for a disk 
with an inner hole, radius ri, is convenient using this 
form, because the integral can be written as the sum of 
two contributions from scattering angles between 92 and 
9i and between tt — 9i and Tr — 92- The Stokes parameters 
are then calculated according to 

Iib)^Cp{b) + Cs{b), 
Q{b) = Cpib)-Cs{b), 
U{b)^0. (7) 

According to this convention, the scattering plane (the 
disk) defines the +Q axis. If the grains are asymmet- 
ric and their orientations are not random, or the disk 
is not exactly edge on then U 7^ 0. However, there is 
no evidence from our current observations that this as- 
sumption is violated. As the disk is optically thin, mul- 
tiple dust components can be represented by summing 
the contribution from different grain populations; such 
components may include grains of various sizes or com- 

4.1. A Henyey-Greenstein Model 

We begin with a simple grain scattering model that il- 
lustrates both the feasibility of simultaneous fitting of 
the surface brightness and degree of polarization and 
the nature of the resultant constraints. Although any 
phase function can be written as a sum of Legendre 
polynomials, our goal is to construct a model with the 
minimum number of free parameters. Therefore, we 
adopt the empirical Henyey-Greenstein function as a con- 
venient approximation to sc attering by small particles 
l(Henvev fc GreensteinI 11941'). We assume that the az- 
imuthal dependence of polarization obeys a Rayleigh law, 
with peak linear polarization parameterized by pmax < 1- 

Polarization of AU Mic 


The corresponding elements of the intensity scattering 
matrix are given by 


2 ^ ^ 4^(l+g2_2gcos( 


with — 1 < g < 1, and 


1^1 1 


\si\' + \s2n 




where for consistency with Eq. ^ the PA of the elec- 
tric vector is measured from the scattering plane. For 
Rayleigh scattering < Pmax < 1, and Q < 0. In the 
case of Mie scattering from dielectric spheres, Eqs. ^ 
and are a satisfactory approximation for grains with 
a; < 1. This recipe cannot describe a second peak at 
9 — TT associated with enhanced backscatter or give po- 
larization parallel to the scattering plane. Nonetheless, 
it has several desirable features: 1) there are only two ad- 
justable grain parameters — the other model parameters 
are the normalization, the inner and outer disk radii, 
and the power law slope of the radial dust distribution; 
2) the computational simplicity of evaluating Eqs. IHland 
13 This is a consideration as performing simultaneous, 
non-linear least-squares fits requires multiple (~ 10'^) nu- 
merical evaluations of the integral in Eq. ©. 

Figure shows the least-squares fit to the Henyey- 
Greenstein model and Tableniists the fit parameters and 
goodness-of-fit, xt- The fit was found using Craig Mark- 
wardt's constrained, non-linear least squares program, 
MPFIT^, which is implemented in the IDL programming 
language^. As we are fitting both the surface brightness 
and the polarization, the results of the fit depend on the 
relative weights attributed to each data set. For the cur- 
rent observations the errors in the surface brightness are 
smaller than for the fractional polarization. Therefore, 
we have assigned a minimum 10% fractional uncertainty 
to the surface brightness to refiect our prejudice that the 
polarization data carry significant information regarding 
the nature of the grains. This assumption is justified 
by noting that NE and SW wings of the disk are not 
identical — these local variations in the dust column, e.g., 
caused by density waves or the injection of fresh mate- 
rial, cannot be fit by our simple power law model, and 
therefore these deviations should not contribute to xt- 

The fit achieves xt = ^■'^ with i/ = 146 degrees of 
freedom, which is gratifyingly good given that there are 
only six parameters. The model accurately reproduces 
the shape of the surface brightness profile, the steep rise 
in polarization between 20 and 40 AU, and is consistent, 
within the errors, with the leveling off at p ~ 0.40 beyond 
this point. Key aspects of the fit include highly polariz- 
ing grains, Pmax — 0.53 ±0.03, strong forward scattering, 
g = 0.68 ± 0.01 and an inner hole at ri = 38 ± 0.5 AU. 
The quoted errors are only the formal errors and should 
be treated with some caution. For example, the best-fit 
value of r2 is biased by the fact that our last data point 
lies at 80 AU. If the outer radius is held fixed at 200 AU, 
then xt increases to 1.76, which is unacceptable only at 
the 1-cr level and perhaps indicative that a single grain 
population cannot account for the scattered light from 

http: //cow .physics .wise . edu/$\sim$craigm/ 

http: //www. ittvis . comP 

the entire disk. Back scattering, which is typical of parti- 
cles with X > 1 can be described by a simple extension of 
Eq. © to a two component Henyey-Greenstein function 
H{9) = (1 - B)H{0,gi) + BH{6,g2), where < 51 < 1 
and — 1 < 52 < 0. The introduction of two extra scat- 
tering parameters does not achieve an improved fit, and 
therefore we find no evidence for enhanced backscatter. 

These disk models teach us that the radial dust dis- 
tribution and the phase function are covariant if only 
Stokes / is available. Evidently, a uniform disk with a 
high degree of forward scattering can mimic a disk with 
a steep decline in grain opacity that is combined with 
more isotropic scattering. The results of analyses that 
adopt a specific radial profil e, e.g., by fixing /9, must be 
interpreted accordingly fe.g.. lGolimowski et alJl20(M on 
(3 Pic). For parameterized grain properties, e.g., Eqs. 
© and © this covariance remains. When a physical 
scattering model is adopted, which ties together g and 
Pmax, and / and Q are fitted simultaneously, then this 
degeneracy is broken. 

4.2. A Zodiacal Dust Model 

Now that we have shown that a simple model can re- 
produce the observations of the AU Mic disk we can ask 
whether grains with the optical scattering properties con- 
sistent with experime ntal studies of Solar System dust 
work too. We adopt iHond l)1985() 's description of the 
scattering phase function inferred from the observed an- 
gular variation of the surface brightness and polarization 
of the Zodiacal light. Hong's formulation is convenient 
because it represents the scattering characteristics of in- 
terplanetary particles as a three component linear com- 
bination of three Henyey-Greenstein functions. 

T his exa mple reproduces the calculation that 
IArtvmowic3 l|1997j) used to describe the polarization 
signature of j3 Pic. There are now only four free param- 
eters: the normalization of the vertical optical depth, 
the radii of the inner and outer holes and the power 
law index of the radial density distribution. Figure El 
shows that this model fits the surface brightness profile 
well, but fails to provide an adequate description of the 
measured polarization. Chi-squared for the combined 
data set is 6.29, which can be rejected with high confi- 
dence (>99%). The Zodiacal dust model cannot explain 
the steep rise in polarization over the inner disk (20-50 
AU) and it cannot account for the high polarization in 
the outer disk. The radial extent of the disk is similar 
to that of the Henyey-Greenstein model with an inner 
hole at 37 ± 1 AU. The disk terminates at r2 = 90 AU. 
Since the scattering asymmetry parameter is fixed at 
g — 0.4 the radial opacity distribution is steeper than in 
the Henyey-Greenstein model. A satisfactory fit to the 
polarization data cannot be found, even if we set the 
weights for the surface brightness to zero. 

The lessons from these results are twofold. First, sur- 
face brightness data alone are insufficient to constrain 
grain optical properties and their radial distribution. 
Second, particles with the optical properties of Zodia- 
cal grains cannot explain the polarization signature of 
AU Mic. Thus, the conclusions of the qualitative discus- 
sion in t|3.2l are not fully borne out: simply changing the 
radial dust distribution does not explain the difference 
between (3 Pic and AU Mic. 


Graham et al. 

10 100 

b [AU] 

Fig. 6. — Simultaneous fits to the surface brightness profile (top) and the degree of linear polarization (bottom). Three different models 
are shown (see Table 1 for details). The grains in the best fit model (solid line) are porous (91%) water ice with xl = 1-6- The dash-dot 
line is a single-component Henyey-Greenstein model = 1-7. The dashed line is a Zodiacal dust model. The Zodiacal dust cannot account 
for the observed polarization fraction (x^ = 6.3). The phenomenological Henyey-Greenstein and the physical porous models can explain 
both observations. They have in common a high degree of forward scattering and polarization. Models where the dust grains are solid 
yield unacceptable fits. The error bars in the lower panel include the systematic uncertainty in the degree of polarization. 

Choosing an alternate type of Solar System dust, e.g., 
cometary grains, does not significantly improve the dis- 
crepancy between the model and the polarization data 
because cometary dust does not have sufficiently high 
peak linear polarization. The dusty comets, of which 
Comet 1996 B2 (Hyakutake) is typical, show little dis- 
persion in their polarizing properties. For example, the 
maximum degree of linear polarization of Comet 1996 
B2 (Hyakutake) was observed to be 0.24 and 0.26 at 484 
and 6 84 nm, respectively, at a phase angle near 94 de- 
grees (jKiselev fc Velichkol[l99^ . 

4.3. Dielectric Spheres: Mie Theory 

Several debris disk studies have used Mie theory 
to evaluate the compl ex elements of the amplitude 
scattering matrix ( e.g., iVoshchinnikov k. Krugell [19991 
IKrivova et al.l l2000() . Solar System dust particles are 
not spheres, and computing the cross sections u sing Mie 
theory may be misleading (cf. lGustafsonlll994|) . How- 
ever, Zodiacal and cometary dust have too low a value of 
peak linear polarization to be consistent with the AU Mic 
data. As polarization efficiency increases with decreasing 
particle size, with Raylcigh scatterers representing the 
limiting case, it is worth investigating whether small di- 
electric spheres composed of common astrophysical ma- 
terial can be invoked. Perhaps erosion in AU Mic's disk is 

so severe that the grains have been ground down to their 
constituent interstellar precursors? Since internal grain 
structure can be neglected as a; — + the Mie approx- 
imation should be sufficiently accurate to explore this 

Mie models with small grains can explain the ob- 
served polarization. Adopting astronomical "silicate" 
(m = 1.65 — O.Oli) as the grain material yields a joint 
fit that is better than the Zodiacal dust model, but the 
best value of = 4.4, is clearly unacceptable at a high 
level of confidence (> 99%). Other parameters are fisted 
in Table Most of the contributions to are from 
residuals relative to the surface brightness profile, which 
are attributable to grain scattering that is too isotropic. 
Only a narrow range of spherical particle sizes come close 
to approximating the data because the joint fit simulta- 
neously constrains the phase function and the maximum 
polarization — quantities that vary rapidly with particle 
size. The best fit size parameter is 2: = 1.63 ± 0.01 for 
astronomical "silicate", or a = 0.16 /im. Adopting dirty 
water-ice grains (m = 1.33 — O.Oli) reduces signifi- 
cantly, but not to an acceptable level. Organic refractory 
material (m = 1.98 — 0.28«; Li & Greenberg 1997) fares 
worse than either rock or ice. 

The conclusion that xt varies significantly with choice 
of n, the real part of the refractive index, suggests that it 

Polarization of AU Mic 


Disk Models 







Porous water ice^ 







(an 7^ 






1. / 


(n ni ~i 











Water ice'' 
































SS Zodi'' 










(a) Maxwell-Garnett/Mie model for porous water ice with m = 
1.33 - O.Oli. The best fit porosity is 0.91 ± 0.09. The best fit grain 
size corresponds to i: = 3.26 ± 0.15 (620 it 30 nm diameter) The 
peak linear polarization pmax asymmetry parameter g are derived 

(b) Formal l-a uncertainties are indicated in parentheses. Derived 
parameters that are not model parameters have a blank in the 
second row. The use of ". . ." implies that the corresponding pa- 
rameter is fixed. 

(c) Single component Henyey-Greenstein model. 

(d) Single component Henyey-Greenstein model with r2 fixed. 

(e) Single particle Mie model with best fit size parameter, x = 
2.13 it 0.01 for solid "dirty ice" grains (m = 1.33 + O.Oli). 

(f) Interstellar dust model (White 1979). 

(g) Single particle Mie model with best fit size parameter, x = 
1.63 it 0.01 for solid "silicate" grains (m = 1.65 — O.Oli). 

(h) Solar System Zodiacal three-component Henyey-Greenstein 
dust model (Hong 1985). 

should be adopted as a fit parameter. Such models have 
six free parameters — the same as the Henyey-Greenstein 
model of tl4.1l A satisfactory fit, with = 1.6, is 
achieved for n — 1.03±0.03 and x — 3.25±0.18 (solid line 
in Fig. 0). Based on the value of x^, this low-index Mie 
fit is slightly better than the Henyey-Greenstein model, 
and this model is consistent with the disk extending be- 
yond 200 AU. While our choice of the complex part of the 
refractive index is somewhat arbitrary, making the grains 
more or less absorbing does not qualitatively change our 

Using our best fit model we also investigate whether 
or not the inner disk (r < ri) is dust free. By adding 
a parameter that describes an inner hole with constant 
vertical depth, we find that r^(r < ri) < 0.003 T±{ri) 
(99% confidence). The inner disk is devoid of micron- 
sized grains, which according to the iStrubbe fc Ghiand 
l)2005j) model means that collisions dominate, i.e., this is 
a "Type B" disk where grains that are dragged inward 
by corpuscular and Poynting-Robertson drag undergo a 
destructive collision. 


Figurel^and the corresponding fit parameters in Table 
n demonstrate that a variety of radial grain distributions 
can account for the observed surface brightness, although 
a large inner hole with radius of 40-50 AU is common to 
all models. Taken together, the surface brightness and 

degree of linear polarization narrow down the range of 
acceptable grain optical properties. Inspection of Figure 
[7| shows that a combination of strong forward scatter- 
ing and a high polarizing efhciency, with a Rayleigh-like 
sin^6'/(l + cos^ 6*) angular variation is sufficient to de- 
scribe these data. Our analysis implies with high confi- 
dence that the constituent dust grains exhibit high max- 
imum linear polarization {pmax > 0.50) and strong for- 
ward scattering {g > 0.7). 

Plots of the phase function and polarization fraction 
versus phase angle shown in Figure [T] help highlight the 
common features of the statistically acceptable Henyey- 
Greenstein and low- index Mie models, and how their 
scattering properties differ from conventional rocky or 
ice-grain models. Large rocky or icy grains typically have 
strong forward scattering, but polarization which oscil- 
lates with phase angle. Therefore, Q tends to average 
to zero along a line of sight that integrates over a range 
of scattering angles (refer to H3.1(l . As the conventional 
Mie fit shows, small grains are good polarizers, but scat- 
ter too isotropically to be consistent with the data. The 
low index Mie model suggests a physical scenario that 
combines strong forward scattering and high polariza- 
tion. ^^^^^ 

Application of the Clausius-Mossotti relation (IJacksonI 
1962) implies that such a low value of n must be asso- 
ciated with a very porous medium — the only terrestrial 
analog that comes to mind is silica aerogel. Aerogels 
are transparent, highly porous materials of low density, 
ranging from 0.05 to 0.15 g cm~'^, with a corresponding 
refractive index of 1.01 to 1.04, respectively. Aerogel has 
porosity on a micron scale and is composed of individual 
silica grains with diameters of ~ 10 nm, which are linked 
in a highly porous dendritic backbone. Although aerogel 
is produced in a process that is unlikely to occur in an 
astrophysical setting — hydrolysis of methyl silicate in the 
presence of a solvent (ethanol) that is subsequently evap- 
orated at high temperature and pressure — the compari- 
son is not entirely frivolous. The appearance of aerogel 
is often characterized as "solid blue smoke", because to 
a good approximation the scattering is Rayleigh scatter- 
ing (Kami uto et alJll99^ . Thus, aerogel is an example 
of a bulk material that interacts with electromagnetic 
radiation in a way that is determined by its microscopic 

Allowing the real part of the refractive index to vary as 
described in i |4. 31 while holding the imaginary part fixed 
violates the Kramers-Kronig relation. An approach that 
has a better physical basis is to use an effective medium 
theory to compute the optical behavior of a porous parti- 
cle describ ed as vacuum m atrix (n = I) with embedded 
inclusions ijKruegell 12003*) . Using the Maxwell-Garnett 
rule we can choose a refractive index for the bulk mate- 
rial and use the grain porosity as a model parameter. The 
best-fit porous grain model (see Fig. Eland the first line 
of Table is practically identical to the variable index 
fit, yielding essentially the same structural parameters. 
The grain porosity is 91-94%, depending on whether we 
assume that the matrix from which our grains are made 
is ice or rock. 


Graham et al. 

Fig. 7. — Normalized phase function (top) and polarization fraction (bottom) versus scattering angle. The heavy solid line is inferred 
from the best fit single component Henyey-Greenstein model with scattering asymmetry parameter g = 0.68. Here polarization denotes 
—Q/I, thus negative polarization indicates that the electric field is oriented parallel to the scattering plane. On the left are the results of 
Mie calculations for spheres with x = 1, 2 and 3 with conventional m = 1.65 — O.Oli ("silicate"). The best fit single particle size Mie fit 
(Table 1), has x = 1.63 ± 0.01. The inability of a sphere to simultaneously match the phase function and the polarization explains why 
this model can only achieve a poor fit. Small dielectric spheres with x ~ 3 can account for the strong forward scattering, but they cannot 
simultaneously provide a high degree of polarized light perpendicular to the scattering plane. On the right is shown the result when the 
real part of the refractive index is allowed to be a free parameter. Grains with m = 1.033 — O.Oli and x = 3.25 provide a satisfactory fit 
to the data — at least as good as the one component Henyey-Greenstein model. The low index means that the grains must be extremely 
porous (> 90%). 

As our aerogel analogy reminds us, porous materials 
are likely highly anisotropic, and we may not be free to 
assume that we can neglect the effects of nearest neigh- 
bors within the matrix. We therefore examine the results 
of numerical calculations (e.g., discrete-dipole approxi- 
mation and transition-matrix) of light scattering by ag- 
gregates to understand whether or not our interpretation 
of the Mie results in terms of porous grains is credible. 

A lucid exposition of the transition-matrix method 
applied to composite interstellar grains is given by 
llati et af] l)2004j) . Using this method iPetrova et al.l 
()2000|) show results for two instances of silicate (m = 
1.65 — O.Oli) grain clusters consisting of 31 particles or 
"monomers" each with = 1.5. A relatively compact 
aggregate with approximately 70% porosity has 5 = 0.75 
and Pmax = 0.52. The more porous particle (81%) has 
g = 0.75 and pmax — 0.65. In neither case does the 
degree of polarization oscillate with phase angle. These 
clusters have optical properties which make them promis- 
ing analogs of the material inferred to dominate the AU 
Mic disk. 

iKimura et al.l l)2006(l present additional results for 
larger, more porous aggregates. Figure|Hlshows the phase 
function and polarization for a large {xc = 10.2) porous 
(90%) silicate cluster (m = 1.6— O.Oli) composed of 128 
small {xm — 0.9) monomers. This particle has g — 0.84 
and Pmax = 0.82, and has optical properties which make 
it an excellent candidate material for the AU Mic disk. 
Also shown is the corresponding Mic calculation with 
dielectric properties derived using the Clausius-Mossotti 
relation. It is evident that the Mie calculation is only a 
rough approximation — the polarization curve is reason- 
ably well reproduced, and g is overestimated by about 
15%. Although the qualitative conclusion that implicates 
porous grains is secure it seems unlikely that the accu- 
racy of effective medium theories is sufficient, for exam- 
ple, to distinguish between different coagulation schemes 

that are characterized by different porosity. It will be 
necessary to abandon Mie theory in favor of numerical 
modeling of aggregate scattering in the next stage of de- 
bris disk modeling. 

If highly porous aggregate grains explain the polar- 
ization signature of AU Mic's debris disk, and if the 
dust beyond ri ~ 40 AU originates from a "birth 
rin g" of parent bodie s <10 cm in size, as envisioned 
bv lStrubbe fc Chian g (2005), then this porosity may be 
a signature of the agglomeration process whereby in- 
terstellar grains first grew into macroscopic sized ob- 
jects. In the inner Solar System porous particles oc- 
cur naturally in cometary dust, where the sublimation of 
ices leaves a " bird's nest" of refractory organic and sil- 
icate material (jGreenberg fc Hag&,.1990') . Porous grains 
in t he /3 Pic disk may or iginate from cometary activ- 
ity ijLi fc Greenberd 11997(1 . However, the birth ring in 
AU Mic lies safely outside the '^l AU ice sublimation 
point. Based on coUisional lifetime arguments, the size 
of the parent-bodies that suppl y the observed dust in AU 
Mic is in the decimeter range llStrubbe fc Chiangi r2005) . 
Though the existence of larger bod ies that will su ffer 
compaction and restructuring (Blun Tfc WurmI f200Cl() is 
not excluded, they are not the dominant reservoir for 
dust observed at optical and near-infrared wavelengths. 
Evidently, shock compression during attrition of the par- 
ent bodies in the birth ring is not significant. We en- 
vision these bodies as so weakly bound that even the 
most glancing collisions lead to their disruption. Re- 
cent laboratory studies of particle coagulation in the 
proto-solar nebula by ballistic cluster-cluster aggregation 
(|Wurm fc BlumllTgg^ leads to the formation of highly 
(>90%) porous aggregates. Our evidence suggests that 
such a process may have mediated the initial growth of 

Observations of scattered light at a single wavelength 
are primarily sensitive to grains with x ~ 1, and do not 

Polarization of AU Mic 


\ * * • 

AU M'\c HG Fit ■ 

MIe -1 

*"■•••••• • * 

'- - ' \ i \l 

; • Porous Aqgregrote 

50 100 150 

Scattering ongle [degrees] 

Fig. 8. — The phase function and polarization (dots) for a large 
(x = 10.2) porous (90%) silicate particle (77i = L6— ^O.Oli) corn- 
posed of 128 small (x = 0.9) monomers from lKimura e t al. (2006). 
This particle has g = 0.84 and pmax = 0.82. The dashed line is the 
phase function {^g = 0.68) and polarization {pmax = 0.53) for the 
Henyey-Greenstein model which best fits the AU Mic disk. The 
porous grain is more forward scattering and more polarizing than 
required by the AU Mic data, but nonetheless its optical prop- 
erties show that highly porous aggregates constitute an excellent 
candidate for the AU Mic disk. Also shown is the Mie approxima- 
tion for the porous grain with dielectric properties derived using 
the Clausius-Mossotti relation (m = 1.047 - 0.0007i). The Mic 
calculation is a useful first approximation to the polarization, but 
overestimates g by about 15%. 

place strong constraints on the particle size distribution. 
However, preliminary calculations show that the mea- 
sured polarization is consiste nt with a Dohnanyi spec- 
trum l|Fitzgerald et all l|200fiD and L. E. Strubbe 2006, 
private communication). Polarization measurements in 
the UV through the near-infrared could be used to mea- 
sure grain porosity as a function of grain size. 


We have observed the AU Mic debris disk at F606W 
(broad V) with the POLV polarizing filter set in ACS 
high resolution camera aboard HST. The coronagraph 
and PSF subtraction were used to suppress scattered 
light. The disk light is polarized, with the degree of linear 
polarization rising steeply from 0.05 to 0.35 between 20 
and 50 AU and reaching a maximum of about 0.40 within 
80 AU. The inner and outer working limits are set by sys- 
tematic errors in PSF subtraction and declining SNR, 
respectively. The linear polarization is oriented with the 
electric field perpendicular to the disk, which is charac- 
teristic of scattering by optically thin, small grains. 

We have factored systematic errors in the ACS po- 
larization measurements into our uncertainties. For 
the bright, strongly polarized emission between 35 and 
55 AU these errors and not measurement errors dom- 
inate. However, the ACS/HRC polarization calibra- 
tion campaign is on-going, and analysis of the resultant 
data, including a full Mueller matrix descri ption for the 
HRC-KF606W+POLV combination (cf. the 'Hine s et all 
l|2000 ) analysis of NICMOS) and application to this data 
set will improve the reliability and fidelity of these re- 

AU Mic and /3 Pic have different polarization signa- 
tures. /3 Pic shows a shallower gradient and lower peak in 
polarization fraction. We attribute this different to two 
factors: 1) the two disks are probed on different spatial 
scales relative to their inner and outer boundaries; 2) the 
grains in the AU Mic disk have a higher peak linear polar- 
ization than those of /3 Pic. We place limits on the radial 
distribution of grains and their optical scattering prop- 
erties by performing simultaneous fits to the observed 
surface brightness and the degree of polarization. These 
fits show that the inner boundary of the AU Mic disk 
is located between 40-50 AU, and the dust component 
which is responsible for the strong linear polarization ex- 
tends to 100-150 AU. The uncertainty occurs primarily 
because the inferred spatial structure of the disk and the 
grain optical properties — the scattering asymmetry fac- 
tor g and the peak linear polarization Pmax — are covari- 
ant if g and Pmax are independent. We can state with 
good confidence that g > 0.7 and Pmax > 0.50. The in- 
ner disk is virtually free of micron-sized grain, and Type 
B conditions prevail (cf. Strubbe & Chiang 2005). 

This combination of optical properties occurs naturally 
in porous media. Once we adopt a physical description 
for the electromagnetic properties of the scatterers, g and 
Pmax are not independent and the covariance with the 
radial dust distribution is greatly reduced. Our best fit 
physical model, which invokes Mie theory and low index 
grains, implies that the inner regions of the AU Mic disk 
(< 50 AU) are de pleted of small grains. T his finding 
lends support to the lStrubbe fc Chiand l|2005!) birth ring. 

The best fit, porous grain model using Mie theory 
and the Maxwell-Garnett rule implies a grain porosity 
of 91-94%, depending on whether the grain matrix is 
ice or rock. Porous grains are a natural consequence of 
particle growth. However, the accuracy of the effective 
medium theory, which we used to convert the dielectric 
constant into a porosity, is probably insufhcient to favor 
one growth mechanism over another, e.g., cluster-cluster 
vs. cluster-particle agglomeration. Better approxima- 
tions for calculating the optical properties of clusters, 
e.g., the discrete-dipole approximation or the transition 
matrix method must be employed. 

We thank John Biretta for help with understanding 
the polarization calibration ACS. This work is based on 
observations with the NASA/ESA Hubble Space Tele- 
scope obtained at the Space Telescope Science Institute 
(STScI), which is operated by the Association of Uni- 
versities for Research in Astronomy. Support for Pro- 
posal number GO-10228 was provided by NASA through 
a grant from STScI under NASA contract NAS5-26555. 
This work has been supported by the National Science 
Foundation Science and Technology Center for Adap- 
tive Optics, managed by the University of California 
at Santa Cruz under cooperative agreement No. AST- 
9876783. B.C.M. acknowledges support from the Na- 
tional Research Council of Canada. 


Artymowicz, P. 1997, Annual Review of Earth and Planetary Augereau, J. C. & Beust, H. 2006, ArXiv Astrophysics e-prints 

Sciences, 25, 175 


Graham et al. 

Aumann, H. H., Beichman, C. A., Gillett, F. C, de Jong, T., 

Houck, J. R., Low, F. J., Neugebauer, G., Walker, R. G., & 

Wesselius, P. R. 1984, ApJ, 278, L23 
Backman, D. E. & Paresce, F. 1993, in Protostars and Planets III, 

ed. E. H. Levy & J. L Lunine, 1253-1304 
Barrado y Navascues, D., Stauflfer, J. R., Song, L, & Caillault, J.-P. 

1999, ApJ, 520, L123 
Biretta, J. & Kozhurina-Platais, V. 2004, ACS Polarization 

Calibration— II. The POLV Filter Angles: ACS 2004-10 

(Baltimore: Space Telescope Science Institute) 
Biretta, J., Kozhurina-Platais, V., Boffi, F., Sparks, W., & Walsh, 

J. 2004, ACS Polarization Calibration — I. Introduction and 

Status Report: ACS 2004-09 (Baltimore: Space Telescope Science 


Blum, J. & Wurm, G. 2000, Icarus, 143, 138 

Chandrasckliar, S. 1960, Radiative transfer (New York: Dover, 

Chen, C. H., Patten, B. M., Werner, M. W., Dowell, C. D., 
Stapelfeldt, K. R., Song, I., StaufTer, J. R., Blaylock, M., Gordon, 
K. D., & Krause, V. 2005, ApJ, 634, 1372 

Evans, D. S., Menzies, A., & Stoy, R. H. 1957, MNRAS, 117, 534 

Fitzgerald, M. P., Kalas, P. G., Duclicne, G., Piiite, C., & Graham, 
J. R. 2006, to be submitted to ApJ, 000, 000 

Ford, H. C., Clampin, M., Hartig, G. F., ct al. 2003, in 
Future EUV/UV and Visible Space Astrophysics Missions and 
Instrumentation. Edited by J. Chris Blades, Oswald H. W. 
Siegmund. Proceedings of the SPIE, Volume 4854, pp. 81-94 
(2003)., ed. J. C. Blades & O. H. W. Siegmund, 81-94 

Gledhill, T. M., Scarrott, S. M., & Wolstencroft, R. D. 1991, 
MNRAS, 252, 50P 

Golimowski, D. A., Ardila, D. R., Krist, J. E., Clampin, M., Ford, 
H. C, Illingworth, G. D., Bartko, F., Bem'tez, N., Blakeslee, J. P., 
Bouwens, R. J., Bradley, L. D., Broadhurst, T. J., Brown, R. A., 
Burrows, C. J., Cheng, E. S., Cross, N. J. G., Demarco, R., 
Feldman, P. D., Franx, M., Goto, T., Gronwall, C, Hartig, G. F., 
Holden, B. P., Homeier, N. L., Infante, L., Jee, M. J., Kimble, 
R. A., Lesser, M. P., Martel, A. R., Mei, S., Menanteau, F., 
Meurer, G. R., Miley, G. K., Motta, V., Postman, M., Rosati, 
P., Sirianni, M., Sparks, W. B., Tran, H. D., Tsvetanov, Z. I., 
White, R. L., Zheng, W., & Zirm, A. W. 2006, AJ, 131, 3109 

Greenberg, J. M. & Hage, J. I. 1990, ApJ, 361, 260 

Gustafson, B. A. S. 1994, Annual Review of Earth and Planetary 
Sciences, 22, 553 

Heiles, C. 2000, AJ, 119, 923 

Henyey, L. G. & Greenstein, J. L. 1941, ApJ, 93, 70 
Hines, D. C, Schmidt, G. D., & Schneider, G. 2000, PASP, 112, 

Hong, S. S. 1985, A&A, 146, 67 

lati, M. A., Giusto, A., Saija, R., Borghese, F., Denti, P., Cecchi- 

PestcUini, C, & Aiello, S. 2004, ApJ, 615, 286 
Jackson, J. D. 1962, Classical Electrodynamics (Classical 

Electrodynamics, New York: Wiley, 1962) 
Kalas, P., Graham, J. R., & Clampin, M. 2005, Nature, 435, 1067 

Kalas, P., Liu, M. C, & Matthews, B. C. 2004, Science, 303, 1990 
Kamiuto, K., Saitoh, S., & Tokita, Y. 1993, J. Quant. Spectrosc. 

Radiar. Transfer, 50, 293 
Kimura, H., Kolokolova, L., & Mann, I. 2006, A&A, 449, 1243 
Kiselev, N. N. & Velichko, F. P. 1998, Icarus, 133, 286 
Kozhurina-Platais, V. & Biretta, J. 2004, ACS Polarization 
Cahbration — III. Astrometry of Polarized Filters: ACS 2004-11 
(Baltimore: Space Telescope Science Institute) 
Krist, J. E., Ardila, D. R., Golimowski, D. A., Clampin, M., Ford, 
H. C, Illingworth, G. D., Hartig, G. F., Bartko, F., Bem'tez, 
N., Blakeslee, J. P., Bouwens, R. J., Bradley, L. D., Broadhurst, 
T. J., Brown, R. A., Burrows, C. J., Cheng, E. S., Cross, N. J. G., 
Demarco, R., Feldman, P. D., Franx, M., Goto, T., Gronwall, 
C, Holden, B., Homeier, N., Infante, L., Kimble, R. A., Lesser, 
M. P., Martel, A. R., Mei, S., Menanteau, F., Meurer, G. R., 
Miley, G. K., Motta, V., Postman, M., Rosati, P., Sirianni, M., 
Sparks, W. B., Tran, H. D., Tsvetanov, Z. I., White, R. L., & 
Zheng, W. 2005, AJ, 129, 1008 
Kiivova, N. A., Krivov, A. V., & Mann, I. 2000, ApJ, 539, 121 
Kruegel, E. 2003, The physics of interstellar dust (The physics of 
interstellar dust, by Endrik Kruegel. loP Series in astronomy 
and astrophysics, ISBN 0750308613. Bristol, UK: The Institute 
of Physics, 2003.) 
Kundu, M. R., Jackson, P. D., White, S. M., & Melozzi, M. 1987, 
ApJ, 312, 822 

Larwood, J. D. & Kalas, P. G. 2001, MNRAS, 323, 402 
Li, A. & Greenberg, J. M. 1997, A&A, 323, 566 
Liu, M. C. 2004, Science, 305, 1442 

Liu, M. C, Matthews, B. C, Williams, J. P., & Kalas, P. G. 2004, 

ApJ, 608, 526 

Metchev, S. A., Eisner, J. A., Hillenbrand, L. A., & Wolf, S. 2005, 

ApJ, 622, 451 

Pavlovsky, C. 2006, ACS Data Handbook, Version 5.0 (Baltimore: 

Space Telescope Science Institute) 
Petrova, E. V., Jockers, K., & Kiselev, N. N. 2000, Icarus, 148, 526 
Pettersen, B. R. & Hsu, J.-C. 1981, ApJ, 247, 1013 
Saar, S. H., Martens, P. C. H., Huovehn, J., & Linnaluoto, S. 1994, 

A&A, 286, 194 

Schmidt, G. D., Elston, R., & Lupie, O. L. 1992, AJ, 104, 1563 
Strubbe, L. E. & Chiang, E. I. 2005, ArXiv Astrophysics e-prints 
Tamura, M., Pukagawa, M., Kimura, H., Yamamoto, T., Suto, H., 

& Abe, L. 2006, ApJ, 641, 1172 
van de Hulst, H. C. 1981, Light scattering by small particles (New 

York: Dover, 1981) 
Voshchinnikov, N. V. & Krugel, E. 1999, A&A, 352, 508 
Wolstencroft, R. D., Scarrott, S. M., & Gledhill, T. M. 1995, 

Ap&SS, 224, 395 
Wurm, G. & Blum, J. 1998, Icarus, 132, 125 

Zuckerman, B., Song, I., Bessell, M. S., & Webb, R. A. 2001, ApJ, 
562, L87